I did the same kind of thing as last week, with the following changes:
- Used 0.7Alpha3-rc1
- Traced a first-time publish of a small share -- should be data-
intensive due to all the uploading
- Correlated the twisted.log file to the TCPDump output to see what
was attempted in each connection
- Drew some colorful graphs (ooh, pretty) for visualization
See updated: http://wiki.osafoundation.org/bin/view/Journal/
PublishConnectionAnalysis.
Really quick summary: the lowest hanging fruit here is reusing
connections, avoiding actions that make the server drop connections
(or having the server not drop connections by fixing the 501 response
bug), and finding out what's going on while the TCP connection is
live but the SSL handshake isn't started yet (a total of 7.5 seconds
spent in this state).
Lisa
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
